GaN/InGaN-based Light-Emitting Diodes (LEDs), known as “Blue LEDs,” have a promising future. Practical applications for these GaN/InGaN-based LEDs have been expanding to include such products as mobile phone key-pads, LCD backlights, traffic lights, commercial signs, automotive lights, outdoor full-color display panels, household illuminative devices, and others. In these and other applications, these high-brightness LEDs may replace conventional light sources such as incandescent and fluorescent lights. Blue LEDs are characterized by high light output at lower energy input than conventional light sources (energy saving, high efficiency) and a longer working life. Their high performance and reliability shows promise for their successful replacement of conventional light sources; however, there is a need to improve current LED designs to overcome currently-known limitations and inherent drawbacks. Better and more precise manufacturing techniques help advance blue LED design by cutting waste, increasing yields, and allowing more advanced and complex or improved designs to emerge, advancing the technology through more flexibility in Design for Manufacturability (DFM). Such improved manufacturing techniques simplify and reduce the cost of their manufacture.
Blue LED's may be fabricated by depositing GaN/InGaN layer(s) on a sapphire substrate. Once the LED devices have been fabricated, the wafer is separated into individual dies. One current die separation process involves the following steps. First the sapphire wafer is thinned to less than 100 μm in thickness by grinding and lapping the backside of the wafer. Next the wafer is mounted to dicing tape and then scribed along the streets between the die by means of a diamond scribe tip or UV laser beam. Finally, the wafer is fractured along the scribe lines by means of a fracturing tool. After fracturing, the dicing tape is stretched so as to physically separate the die from one another so that subsequent automated pick and place operations can be performed. This process is referred to as “scribe and break” die separation.
A major cost of LED fabrication is the sapphire thinning and the scribe-and-break operation. A process known as LED lift-off can dramatically reduce the time and cost of the LED fabrication process. LED lift-off may eliminate wafer scribing by enabling the manufacturer to grow GaN LED film devices on the sapphire wafer, for example, and then transfer the thin film device to a heat sink electrical interconnect. In this process, the laser beam profile fires through the back of a sapphire wafer to de-bond the GaN LED device and transfer it to a substrate where it can then be packaged onto a heat sink and/or optical reflector. Using special wafers, the sapphire growth substrate may possibly be re-used, and the cost of LED fabrication can be reduced. Additionally, this approach is fast, delivering increased LED light output, and has low operating costs due to low stress on the UV laser.
Current designs of GaN LEDs have inherent limitations that hamper efforts to improve performance and reliability. The designs have also been associated with electrostatic discharge problems. As shown in FIGS. 1A and 1B, a blue LED 10 may include multiple InGaN and GaN based layers 12a, 12b, 12c which are hetero-epitaxially grown on a silicon carbide or a sapphire wafer substrate 14. Since the sapphire wafer is a natural insulator, current is supplied by a horizontal electrode configuration. Due to the high resistance of the p-GaN layer 12a, a thin film of Ni/Au 16 is deposited over the p-GaN to promote current dispersion spreading. However, there are some drawbacks associated with the horizontal configuration.
First, the Ni/Au film 16 absorbs a substantial portion of the LED light output. The Ni/Au film 16 is very thin (usually less than 100 Å), in order to make it transparent to LED light, since it has limited transmittance to the emitting light. Approximately 25% of the light emitted by the LED itself is absorbed by the Ni/Au film 16. Furthermore, a significant percentage of the emitted light is lost in transmission through the sapphire. Some of light directed towards the sapphire substrate 14 is reflected to the front surface due to the difference in refractive indices between the sapphire wafer and its surroundings. The Ni/Au thin film 16 absorbs the majority of this reflected output light as well.
Secondly, the Ni/Au film 16 is sensitive to moisture, resulting in performance degradation over time. To maintain the film's transparency, thin Ni/Au is deposited by metal evaporation, and then heat-treated in an ambient air or an O2 environment. The Ni/Au film 16 forms an oxidized compound, NiOx with an Au-rich structure. When moisture penetrates to the oxide film over long-term operation, the LED device 10 will be damaged.
Third, the Ni/Au film 16 experiences a degradation in the performance efficiency of the InGaN MQW light-emitting layer 12b due to a current crowding effect. Since the current spreading Ni/Au film 16 has lower resistance than the n-GaN layer 12c, the current may crowd in the region 18 near the n (−) electrode 20 (see FIG. 1A). Thus, the phenomenon of current crowding may prevent homogeneous use of the active InGaN area, resulting in low efficiency of light output and low reliability due to uneven use of the active area.
Fourth, the horizontal-electrode configuration may create the effect of a current bottleneck, resulting in low reliability. The current supplied through the p (+) electrode 22 spreads across the Ni/Au film 16, and flows from p-GaN 12a through InGaN 12b to n-GaN 12c. Since the n (−) electrode 20 is horizontally located at the n-GaN 12c, the current is bottlenecked in the area 24 at the electrode 20 (FIGS. 1A and 1B).
A LED structured with a vertical electrode configuration overcomes many of the drawbacks of the horizontal LED structure. As shown in FIG. 2, an LED 30 with a vertical structure involves a transfer of GaN layers 32a, 32b, 32c from the sapphire substrate to a conductive substrate 34, such as a silicon wafer. The vertical electrode configuration may eliminate the Ni/Au film, which substantially increases light output. The vertical structure allows the deposition of a metal reflection layer 36, which minimizes light loss through the sapphire in the horizontal structure. The vertical structure also improves reliability and performance by reducing or eliminating the current crowding and bottle neck. A factor in constructing the vertical LED structure is the successful lift-off process of the GaN layer from the epitaxial sapphire wafer to the conductive silicon wafer.
One example of the construction of a high-brightness vertical LED is show in FIG. 3. First, GaN layers 32a, 32c are deposited onto a sapphire wafer 38. After a metal thin-film reflector 36 is deposited on the p-GaN, then a Si substrate, or any other conductive substrate 34 (including GaAs substrate and thick metal films) is bonded over the metal thin-film reflector. The sapphire wafer is removed by UV-laser lift-off, as described below. The n (−) electrode is deposited on the n-GaN layer and the p (+) electrode is deposited on the Si wafer. Since the n-GaN layer has lower resistance than the p-GaN layer, the thin Ni/Au film is no longer needed. Current is therefore more evenly spread without crowding or a bottleneck effect. Elimination of the troublesome Ni/Au thin film results in an increase in performance and reliability of LEDs with the vertical structure.
The vertical structure may be created using a UV-laser lift off process. One approach to UV-laser lift-off involves the selective irradiation of the GaN/Sapphire interface with a UV laser pulse, utilizing the absorption difference of UV light between the GaN (high absorption) thin film layers and the sapphire substrate. Commonly, the GaN layers are hetero-epitaxially grown on a sapphire wafer. To facilitate GaN crystal growth, a buffer layer may be deposited at a relatively low temperature, around 300° C. While the buffer layer helps to grow the GaN layer at a high temperature, the buffer layer contains a very high density of various defects due to a large lattice mismatch. The crystal defects, such as dislocations, nanopipes and inversion domains, elevate surface energy which consequently increases absorption of incident UV light. The incident laser beam for the lift-off process carries an energy density well below the absorption threshold of the sapphire wafer, allowing it to transmit through without resulting in any damage. In contrast, the laser energy density is high enough to cause photo-induced decomposition at the interface, which allows debonding of the interface.
Studies exist regarding the UV laser lift-off process. Kelly et al. demonstrated decomposition of GaN by laser irradiation through transparent sapphire, using a Q-switched Nd:YAG laser at 355 nm. (see M. K. Kelly, O. Ambacher, B. Dalheimer, G. Groos, R. Dimitrov, H. Angerer and M. Stutzmann, Applied Physics Letter, vol. 69 p. 1749, 1996). Wong et al. used a 248 nm excimer laser to achieve separation of ˜5 μm thin GaN film from a sapphire wafer (see W. S. Wong, T. Sands and N. W. Cheung, Applied Physics Letter, vol. 72 p. 599, 1997). Wong et al. further developed the lift-off process on GaN LED using a 248 nm excimer laser (see W. S. Wong, T. Sands, N. W. Cheung, M. Kneissl, D. P. Bour, P. Mei, L. T. Romano and N. M. Johnson, Applied Physics Letters, vol. 75 p. 1360, 1999). Kelly et al. also demonstrated the lift-off of 275 μm thick, free-standing GaN film using a raster scanning of Q-switched 355 nm Nd:YAG laser (see M. K. Kelly, R. P. Vaudo, V. M. Phanse, L. Gorgens, O. Ambacher and M. Stutzmann, Japanese Journal of Applied Physics, vol. 38 p. L217, 1999). Kelly et al. also reported their difficulty in overcoming extensive fracturing of GaN thick film upon the laser lift-off process, due to high residual stresses from a GaN-sapphire wafer. Id. In this study, the authors had to heat the GaN/sapphire wafer to 600° C., but they could not completely offset the fracturing problems caused by the residual stresses.
In spite of the advantages from UV-laser lift-off, GaN LED manufacturing has been limited due to poor productivity caused by low process yield. The low yield is due in part to high residual stresses in a GaN-sapphire wafer, resulting from a Metal-Organic Chemical Vapor Deposit (MOCVD) process. The MOCVD process requires an activation temperature of over 600° C. As shown in FIG. 4A, GaN and InGaN layers 32 are deposited on a sapphire wafer 38 by the MOCVD process. Since there is substantial difference in coefficients of thermal expansion (CTE) between the GaN (5.59×10−6/° K) and the sapphire (7.50×10−6/°K) (see Table 1), high levels of residual stresses exist when the GaN/sapphire wafer cools down to ambient temperature from the high temperature of the MOCVD process, as shown in FIG. 4B. The residual stresses include compressive residual stresses 40 on the GaN and tensional residual stresses 42 on the sapphire.
TABLE 1Various material properties of GaN and sapphire.LatticeLatticeBand GapThermalConst. aConst. cDensityEnergyExpansion ×10−6/MaterialStructure(Å)(Å)(g/cm3)(eV)° KSapphireHexagonal4.75812.9913.979.97.50GaNHexagonal3.189 5.8156.13.35.59
When an incident laser pulse with sufficient energy hits a GaN/sapphire interface, the irradiation results in instantaneous debonding of the interface. Since the incident laser pulse has limited size (usually far less than 1 cm2), it creates only a small portion of the debonded or lifted-off interface. Since surroundings of the debonded area still have high level of residual stress, it creates a concentration of stress at the bonded/debonded border, resulting in fractures at the border. This fracturing, associated with the residual stress, has been one of the obstacles of the UV-laser lift-off process.
Currently, there are different ways to perform laser lift-off processes on GaN/sapphire wafers. One method involves raster scanning of a Q-switched 355nm Nd:YAG laser (see, e.g., M. K. Kelly, R. P. Vaudo, V. M. Phanse, L. Gorgens, O. Ambacher and M. Stutzmann, Japanese Journal of Applied Physics, vol. 38 p. L217, 1999). This lift-off process using a solid state laser is illustrated in FIG. 5A. Another method uses a 248 nm excimer laser (see, e.g., W. S. Wong, T. Sands, N. W. Cheung, M. Kneissl, D. P. Bour, P. Mei, L. T. Romano and N. M. Johnson, Applied Physics Letters, vol. 75 p. 1360, 1999). This lift-off process using an excimer laser is illustrated in FIG. 5B.
Both processes employ raster scanning, as shown in FIG. 6, which involves either translation of the laser beam 44 or the target of the GaN/sapphire wafer 46. A problem associated with the raster scanning method is that it requires overlapping exposures to cover the desired area, resulting in multiple exposures 48 for certain locations. In both of the above methods, the laser lift-off of GaN/sapphire is a single pulse process. The unnecessary multiple exposures in localized areas increase the potential for fracturing by inducing excessive stresses on the film.
As shown in FIG. 7, raster scanning also involves a scanning of the laser beam 44 from one end to the other, gradually separating the GaN/sapphire interface from one side to the other. This side-to-side relaxation of residual stresses causes large differences in the stress level at the interface 50 between the separated and un-separated regions, i.e., the interface between the scanned and the un-scanned area. The disparity in residual stress levels at the interface 50 increases the probability of propagation of Mode I and Mode II cracks. Although the illustrations in FIGS. 6 and 7 are based on a process using a solid state laser, raster scanning of an excimer laser will produce similar results.
Currently, a common size of sapphire wafers is two-inch diameter, but other sizes (e.g., three-inch and four-inch wafers) are also available for the hetero-epitaxial growth of GaN. For a GaN/sapphire wafer, the level of residual stresses varies in the wafer, and compressive and tensile residual stresses may exist together. The existence of the residual stresses may be observed by wafer warping or bowing. When a laser lift-off process relaxes a large area of a continuous GaN/sapphire interface, as described above, a severe strain gradient may be developed at the border between the debonded and the bonded interface. This strain gradient may cause extensive fracturing of the GaN layer.
When a target material is irradiated with an intense laser pulse, a shallow layer of the target material may be instantaneously vaporized into the high temperature and high pressure surface plasma. This phenomenon is called ablation. The plasma created by the ablation subsequently expands to surroundings. The expansion of the surface plasma may induce shock waves, which transfer impulses to the target material. The ablation may be confined in between two materials when the laser is directed through a transparent material placed over the target. During this confined ablation, the plasma trapped at the interface may create a larger magnitude of shock waves, enhancing impact pressures. The explosive shock waves from the confined ablation at the GaN/sapphire interface can cause not only separation of the GaN layer from the sapphire substrate but may also fracture the GaN layer near the laser beam spot (see, e.g., P. Peyre et. al., Journal of Laser Applications, vol. 8 pp.135-141, 1996).
Accordingly, there is a need for an improved method of separating GaN thin films from a sapphire wafer by addressing the problems associated with residual stress, which lead to low yields due to the fracturing of separated film layers. There is also a need for processes that can be extended to any lift-off applications to address one or more of the problems discussed above.